Identifying appropriate reference data models for comparative effectiveness research (CER) studies based on data from...

Area of Science:

• Health Informatics
• Clinical Data Standards
• Comparative Effectiveness Research (CER)

Background:

• Federal endorsement of data standards exists for electronic clinical data exchange.
• A consensus standard for comparative effectiveness research (CER) data has not been established.
• Existing data models show similarities but lack unified adoption in CER.

Purpose of the Study:

• To compare existing clinical data models for their suitability in comparative effectiveness research (CER).
• To evaluate data retrieval and information loss across different data models.
• To identify the most effective data model for CER applications.

Main Methods:

• Qualitative metrics were used to assess data retrieval and information loss.
• Several existing data models were compared: OMOP CDM, BRIDG, CDISC, and FDA Mini-Sentinel.
• The comparison focused on various CER topic areas and data element capture.

Main Results:

• OMOP CDM version 4.0 provided the most detailed data elements for insurance benefit design.
• OMOP and FDA Mini-Sentinel models included standardized vocabularies for semantic interoperability.
• Local data modeling required extensions to standard data models, highlighting a need for flexibility.

Conclusions:

• The OMOP Common Data Model (CDM) demonstrated the broadest alignment with CER objectives.
• Mapping institutional data to assessed models resulted in minimal information loss.
• Enhancements to data dictionaries with local, institution-specific information are necessary for certain CER scenarios.